RESUMO
AIMS: To evaluate the diagnostic accuracy of haemoglobin A1c (HbA1c) in screening for impaired fasting glucose and Type 2 diabetes (T2DM). METHODS: We screened 3904 adults aged 45-70 (mean age 58.6 [standard deviation (SD) 6.9] years, mean body mass index (BMI) 29.9 [SD 4.7]kg/m(2)), with fasting plasma glucose (FPG) and HbA1c as part of a large diabetes prevention programme. We assessed the diagnostic accuracy of HbA1c for predicting impaired fasting glucose (IFG), (defined either as FPG 5.6-6.9 mmol/l, or 6.1-6.9 mmol/l), and T2DM (FPG ≥ 7.0 mmol/l). RESULTS: The prevalences of IFG were 13.8% (FPG 5.6-6.9 mmol/l) and 4.5% (FPG 6.1-6.9 mmol/l) and of T2DM was 2.1%. Using FPG 5.6-6.9 mmol/l as the IFG reference standard, HbA1c of 39-47 mmol/mol (5.7-6.4%) was 63% sensitive and 81% specific, and HbA1c 43-47 mmol/mol (6.1-6.4%) was 21% sensitive and 98% specific, in diagnosing IFG. HbA1c ≥ 48 mmol/mol (6.5%) was 61% sensitive and 99% specific in diagnosing T2DM. Having HbA1c 39-47 mmol/mol (5.7-6.4%), male sex, and body mass index >29.5 together increased the odds of IFG 6.5-fold (95% confidence interval (CI) 5.5-7.8) compared to the pre-test odds. CONCLUSION: Defining 'pre-diabetes' at a lower HbA1c threshold of 39 mmol/mol (5.7%) instead of 47 mmol/mol (6.1%) increases its sensitivity in diagnosing IFG, but current American Diabetes Association definitions of 'pre-diabetes' based on HbA1c would fail to detect almost 40% of people currently classified as IFG. This has implications for current and future diabetes prevention programmes, for vascular risk management, and for clinical advice given to people with 'pre-diabetes' based on fasting glucose data.